Facebook Ads are great, but...
Don’t get me wrong: Facebook Ads have worked fantastically for millions of businesses for years. They deliver results. Arguably, the strongest feature of Facebook ads is Facebook’s impeccable targeting capabilities. There are countless tools available for businesses to improve their ad targeting with Facebook ads; lookalike audiences are among the most effective.
However, the lack of accurate customer lifetime value (CLV) forecasting capabilities is leaving businesses with biased data sets and incorrect assumptions about customers.
These costly faults prevent merchants from realizing the profitability they need to sustain their businesses.
Lookalike audiences are created based on lists of existing customers with the goal of finding new shoppers who are similar to those existing ones. In addition to uploading customer lists, Facebook allows users to include customer lifetime values (CLV) to create value-based lookalike audiences. These value-based lookalike audiences help Facebook improve their targeting so that more valuable prospects are in the audience. However, this does not necessarily improve your bidding strategy. It merely puts more valuable customers in your pool.
Separately, businesses can set a bidding goal to maximize customer value. This bidding strategy is great for any business that is looking for more valuable orders, not just a higher order count. However, there is so much more to unpack about this bidding strategy that Facebook can’t possibly address. In short, Facebook’s bidding maximizes your revenue, not your cash flow or your profit. Remember: you can’t reinvest revenue in your business. You can reinvest profit, so over any lengthy time period, you should always be optimizing for profit.
Customer lifetime value is an extremely important yet underserved metric. Merchants and agencies often do not grasp what CLV truly is. For example, most merchants who chose to create value-based lookalike audiences with CLV data and optimize their bidding strategy for value use historical CLV. Now, this sounds like a great strategy on the surface. Historical CLV is a decent indicator of CLV. However it is far from perfect and in some cases can even be way off the mark. True customer lifetime value is forward looking. CLV is the sum of all future sales from a customer. There are various methods of calculating CLV, however all methods must be forward looking. Historical CLV is the realized value of a customer, whereas CLV is the future value of that customer for the duration of the time that they are a customer. Worst of all, historical CLV values new customers less than it values old customers.
Every single blog post, help guide, or piece of content from an agency that we could get our hands on recommended one of two things: use historical CLV or forecast CLV. Facebook gives a little more advice. Facebook lists the factors involved in CLV and then states merchants should:
“Combine each relevant estimation into a formula appropriate for your business
goals and use the result it produces.” - Facebook
This is in no way, shape, or form sufficient information to enable merchants to forecast CLV. That is because properly forecasting CLV is hard – really hard.
There are many factors that go into CLV not listed by Facebook or any of the agencies. Ecommerce introduces a lot of nontraditional variables to the CLV equation. For example, the usage of discounts and promos is not something that Facebook discusses when considering customer lifetime value, yet it is an essential variable that is crucial in ecommerce. Some merchants have skilled team members with technical backgrounds and can create and refine an in-house formula to calculate CLV. This is an extremely difficult thing to get right, can often lead to costly mistakes, and is a huge time drain. When CLV is calculated improperly, it often would have just been better to use historical data and leave it at that. One reason why formulating a great CLV is so difficult is because it requires forecasting capabilities. To make matters even more complicated, one has to carefully select the variables and data used to calculate CLV.
So why is forecasting so important? Why is historical CLV not sufficient? One word: bias. When CLV is not a forward looking metric, there is a great deal of statistical bias against new customers who are at the beginning of their customer lifetime.
Looking at lifetime value as a historical metric can lead to a variety of incorrect assumptions, all of which lead to Facebook undervaluing new customers, both in audience creation and in bidding. This creates a convoluted picture of what a merchant’s ideal customer looks like, which leads to poor targeting, underbidding on certain high-value customers, and overpaying for low-value customers.
Again, and this cannot be emphasized enough: this is not the fault of Facebook. Facebook’s targeting and bidding systems are state of the art, however Facebook’s AI is only as good as the data you provide it. When Facebook is fed biased data sets, it cannot be expected to perform at peak performance levels; you’ll overpay for bad leads, and you’ll miss the best prospects.
Let’s take a look at a few examples of where this statistical bias becomes problematic. Consider a situation where a brand new customer (Customer A) converts for the very first time. Customer A now has a historical CLV of $40. One of their oldest and most loyal customers (Customer B) has a CLV of $120, so Customer A now appears to have a relatively low CLV. Another customer (Customer C), converts for the second time, purchasing the same product that they ran out of and have come back for more, for $40. They now have a historical CLV of $80. Based on this data, ranked from lowest to highest lifetime value, these customers would rank: Customer A, Customer C, Customer B. When fed this data, Facebook’s AI would consider Customer C to be the shining example of an excellent customer, bidding higher on lookalikes of Customer C than on lookalikes of Customer A or B.
See the problem now? Historical CLV does not paint the full picture. Customer A and C might have low historical lifetime value, but they are new customers. They could have the potential to be top customers, coming back and purchasing products for years to come. In other words, historical data only looks at what customers have already done, while forecasting is estimating what they are going to do. Time is of utmost importance.
Let’s take our example even further. Assume that we have now successfully forecasted each of these customers true CLVs. By forecasted we mean that you are now considering CLV as a forward looking metric, the present value of future cash flows from a given customer. Customer A, B, and C now have lifetime values of $140, $160, and $240.
Ranked from lowest to highest lifetime value, these customers would now rank: Customer C, Customer B, and Customer A. Based on this CLV data, Customer C is now the shining example of an excellent customer. Facebook can now target with more accuracy and bid more on customers who are truly more valuable. In the figures below, you can clearly see that historical CLV painted an incomplete picture of the true value of each customer as the rankings drastically change between historical CLV and forecasted.
Some Closing Thoughts
CLV is an extremely important metric – but efforts to understand and employ it are challenged by the technical difficulties of properly forecasting it.
We live in a world where information overload is commonplace. Getting the attention of customers has never been more difficult. That’s why it is so important to know exactly who your best customers are and how much they are worth to you. This is especially challenging because those metrics are evolving every day.
There comes a point where merchants have monetized their loyal base as much as they reasonably can and now the focus must shift to acquisition. CLV then becomes arguably the most important metric when seeking to acquire more loyal customers. Loyal customers are customers who don’t chronically return products, don’t require large discounts (or any discount) to convert, and who make return purchases for years to come. The only way to truly expand your base of loyal customers with Facebook Ads is to properly forecast CLV, helping Facebook’s AI properly identify your truly loyal customers and find others like them. Merchants who succeed at this will see ROAS and AOV roar higher!
At Pricestack, we’ve cracked the code to provide Facebook the data it needs to optimize ads for value, drastically improving high-value customer acquisition. Pricestack analyzes your shopper data and uses it to create far more accurate customer lifetime value-based lookalike audiences that help Facebook determine what makes someone likely to be a profitable loyal customer rather than just a one-time bargain shopper. Pricestack’s custom pixel event helps Facebook optimize for predicted lifetime profit uplift, not just first-order revenue. This helps Facebook allocate your budget to find prospects who will become profitable, loyal customers. The best part is that installation takes a minute and is risk-free.
Customer lifetime value is the elephant in the room because it’s so hard to calculate and forecast. Optimizing your ads for lifetime profit instead of first-purchase revenue will help you earn more profit from your ads. With Pricestack, you can finally acquire more loyal, profitable customers who will shop from you for years – not days.